三种配体结合残基的识别

The recognition of three LIGAND binding residues

配体结合残基的预测对蛋白质生物功能的理解提供重要的依据,也有助于药物设计和开发。ATP、GTP和NAD三种配体具有相似的化学结构,生物功能也有着紧密的联系,所以我们将ATP、GTP和NAD三种大分子配体作为一个系列,分别对蛋白质中ATP、GTP和NAD配体结合残基进行了识别。建立了ATP、GTP和NAD三种配体结合残基的数据集,分别包含结合残基3838个、1316个和3579个,序列相似性阈值为30%,分辨率阈值为3°A。采用氨基酸组分信息为特征参数的离散增量算法和位点氨基酸保守性信息为特征参数的矩阵打分算法,对以上三种配体结合残基分别进行预测,结果并不理想;之后,我们将二级结构信息和表面可及性信息作为特征参数,并融合离散增量值、矩阵打分值和自相关协方差值,输入支持向量机对结合残基进行预测,得到了较好的预测结果。5交叉检验结果下的ATP、GTP和NAD配体结合残基预测总精度分别为77.4%、82.1%和85.3%;相关系数分别为0.549、0.643和0.702。

The prediction of hgand binding residues can provide important basis for understanding protein biological function, and it is very useful for drug design and development. ATP, GTP and NAD three ligands have similar chemical structures, biological function also have close contact, so we will use ATP, GTP and NAD ligands as a series, and predict ATP, GTP and NAD ligand binding resi- dues in protein. We built the ATP, GTP and NAD three ligand binding residues benchmark datasets, which included binding residues 3838, 1316 and 3579, respectively, sequence identity is below 30% and resolution is better than 3A. Based on aminoacid composition and the amino acid conservative in- formation, we predicted the three kinds of ligands binding residues in proteins, which are characterized by the increment of diversity algorithm and matrix scoring value algorithm. The results are not satis- factory. After that, while with increment of diversity values, matrix scoring values, auto covariance values of physicochemical property, second structure information and surface accessibility information as the parameters for support vector machine, the overall prediction accuracy and MCC of ATP.GTP and NAD achieved 77.4% ,0. 549;82.1% ,0. 643 and 85.3% ,0. 702 by 5--fold cross--validation , re- spectively.